Skip to main navigation Skip to search Skip to main content

Long-tailed Regression with Ensembles for Monocular Height Estimation from Single Remote Sensing Images

  • Technical University of Munich
  • Deutsches Zentrum für Luft- und Raumfahrt (DLR)

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

4 Scopus citations

Abstract

We focus on the long-tailed distribution problem of monocular height estimation from single remote sensing images. The long-tailed distribution results in dramatically large errors for pixels of high buildings. To cope with that, we propose an ensemble network making use of building footprints as the auxiliary information, which leads to improvements of the results. Together with the designed soft label, the network is robust to noises in building footprints. The proposed network is simple, yet practical and able to mitigate the long-tailed effect in monocular height estimation.

Original languageEnglish
Title of host publication2023 Joint Urban Remote Sensing Event, JURSE 2023
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665493734
DOIs
StatePublished - 2023
Event2023 Joint Urban Remote Sensing Event, JURSE 2023 - Heraklion, Greece
Duration: 17 May 202319 May 2023

Publication series

Name2023 Joint Urban Remote Sensing Event, JURSE 2023

Conference

Conference2023 Joint Urban Remote Sensing Event, JURSE 2023
Country/TerritoryGreece
CityHeraklion
Period17/05/2319/05/23

Keywords

  • ensemble
  • long-tailed regression
  • monocular height estimation
  • remote sensing

Fingerprint

Dive into the research topics of 'Long-tailed Regression with Ensembles for Monocular Height Estimation from Single Remote Sensing Images'. Together they form a unique fingerprint.

Cite this